stream: split per-attempt data from clientStream (#1900)

This is pre-work to implementing retry support. Each retry attempt will have its own csAttempt. The fields left in clientStream are the same across all attempts.

call.go

@@ -54,7 +54,7 @@ func invoke(ctx context.Context, method string, req, reply interface{}, cc *Clie
 		}
 		cs := csInt.(*clientStream)
 		if err := cs.SendMsg(req); err != nil {
-			if !cs.c.failFast && cs.s.Unprocessed() && firstAttempt {
+			if !cs.c.failFast && cs.attempt.s.Unprocessed() && firstAttempt {
 				// TODO: Add a field to header for grpc-transparent-retry-attempts
 				firstAttempt = false
 				continue
@@ -62,7 +62,7 @@ func invoke(ctx context.Context, method string, req, reply interface{}, cc *Clie
 			return err
 		}
 		if err := cs.RecvMsg(reply); err != nil {
-			if !cs.c.failFast && cs.s.Unprocessed() && firstAttempt {
+			if !cs.c.failFast && cs.attempt.s.Unprocessed() && firstAttempt {
 				// TODO: Add a field to header for grpc-transparent-retry-attempts
 				firstAttempt = false
 				continue

rpc_util.go

@@ -127,7 +127,7 @@ func (d *gzipDecompressor) Type() string {
 type callInfo struct {
 	compressorType        string
 	failFast              bool
-	stream                *transport.Stream
+	stream                *clientStream
 	traceInfo             traceInfo // in trace.go
 	maxReceiveMessageSize *int
 	maxSendMessageSize    *int

stream.go

@@ -266,29 +266,28 @@ func newClientStream(ctx context.Context, desc *StreamDesc, cc *ClientConn, meth
 		break
 	}
 
-	c.stream = s
 	cs := &clientStream{
 		opts:   opts,
 		c:      c,
 		desc:   desc,
 		codec:  c.codec,
 		cp:     cp,
-		dc:     cc.dopts.dc,
 		comp:   comp,
 		cancel: cancel,
-
+		attempt: &csAttempt{
-		done: done,
+			t:            t,
-		t:    t,
+			s:            s,
-		s:    s,
+			p:            &parser{r: s},
-		p:    &parser{r: s},
+			done:         done,
-
+			dc:           cc.dopts.dc,
-		tracing: EnableTracing,
+			ctx:          ctx,
-		trInfo:  trInfo,
+			trInfo:       trInfo,
-
+			statsHandler: sh,
-		statsCtx:     ctx,
+			beginTime:    beginTime,
-		statsHandler: cc.dopts.copts.StatsHandler,
+		},
-		beginTime:    beginTime,
 	}
+	cs.c.stream = cs
+	cs.attempt.cs = cs
 	if desc != unaryStreamDesc {
 		// Listen on cc and stream contexts to cleanup when the user closes the
 		// ClientConn or cancels the stream context.  In all other cases, an error
@@ -300,7 +299,7 @@ func newClientStream(ctx context.Context, desc *StreamDesc, cc *ClientConn, meth
 			case <-cc.ctx.Done():
 				cs.finish(ErrClientConnClosing)
 			case <-ctx.Done():
-				cs.finish(toRPCErr(s.Context().Err()))
+				cs.finish(toRPCErr(ctx.Err()))
 			}
 		}()
 	}
@@ -311,47 +310,56 @@ func newClientStream(ctx context.Context, desc *StreamDesc, cc *ClientConn, meth
 type clientStream struct {
 	opts []CallOption
 	c    *callInfo
+	desc *StreamDesc
+
+	codec baseCodec
+	cp    Compressor
+	comp  encoding.Compressor
+
+	cancel context.CancelFunc // cancels all attempts
+
+	sentLast bool // sent an end stream
+
+	mu       sync.Mutex // guards finished
+	finished bool       // TODO: replace with atomic cmpxchg or sync.Once?
+
+	attempt *csAttempt // the active client stream attempt
+	// TODO(hedging): hedging will have multiple attempts simultaneously.
+}
+
+// csAttempt implements a single transport stream attempt within a
+// clientStream.
+type csAttempt struct {
+	cs   *clientStream
 	t    transport.ClientTransport
 	s    *transport.Stream
 	p    *parser
-	desc *StreamDesc
+	done func(balancer.DoneInfo)
 
-	codec     baseCodec
-	cp        Compressor
 	dc        Decompressor
-	comp      encoding.Compressor
 	decomp    encoding.Compressor
 	decompSet bool
 
-	// cancel is only called when RecvMsg() returns non-nil error, which means
+	ctx context.Context // the application's context, wrapped by stats/tracing
-	// the stream finishes with error or with io.EOF.
-	cancel context.CancelFunc
 
-	tracing bool // set to EnableTracing when the clientStream is created.
+	mu sync.Mutex // guards trInfo.tr
-
+	// trInfo.tr is set when created (if EnableTracing is true),
-	mu       sync.Mutex
+	// and cleared when the finish method is called.
-	done     func(balancer.DoneInfo)
-	sentLast bool // sent an end stream
-	finished bool
-	// trInfo.tr is set when the clientStream is created (if EnableTracing is true),
-	// and is set to nil when the clientStream's finish method is called.
 	trInfo traceInfo
 
-	// statsCtx keeps the user context for stats handling.
-	// All stats collection should use the statsCtx (instead of the stream context)
-	// so that all the generated stats for a particular RPC can be associated in the processing phase.
-	statsCtx     context.Context
 	statsHandler stats.Handler
 	beginTime    time.Time
 }
 
 func (cs *clientStream) Context() context.Context {
-	return cs.s.Context()
+	// TODO(retry): commit the current attempt (the context has peer-aware data).
+	return cs.attempt.context()
 }
 
 func (cs *clientStream) Header() (metadata.MD, error) {
-	m, err := cs.s.Header()
+	m, err := cs.attempt.header()
 	if err != nil {
+		// TODO(retry): maybe retry on error or commit attempt on success.
 		err = toRPCErr(err)
 		cs.finish(err)
 	}
@@ -359,20 +367,61 @@ func (cs *clientStream) Header() (metadata.MD, error) {
 }
 
 func (cs *clientStream) Trailer() metadata.MD {
-	return cs.s.Trailer()
+	// TODO(retry): on error, maybe retry (trailers-only).
+	return cs.attempt.trailer()
 }
 
 func (cs *clientStream) SendMsg(m interface{}) (err error) {
-	// TODO: Check cs.sentLast and error if we already ended the stream.
+	// TODO(retry): buffer message for replaying if not committed.
-	if cs.tracing {
+	return cs.attempt.sendMsg(m)
-		cs.mu.Lock()
+}
-		if cs.trInfo.tr != nil {
+
-			cs.trInfo.tr.LazyLog(&payload{sent: true, msg: m}, true)
+func (cs *clientStream) RecvMsg(m interface{}) (err error) {
-		}
+	// TODO(retry): maybe retry on error or commit attempt on success.
-		cs.mu.Unlock()
+	return cs.attempt.recvMsg(m)
+}
+
+func (cs *clientStream) CloseSend() error {
+	cs.attempt.closeSend()
+	return nil
+}
+
+func (cs *clientStream) finish(err error) {
+	if err == io.EOF {
+		// Ending a stream with EOF indicates a success.
+		err = nil
 	}
+	cs.mu.Lock()
+	if cs.finished {
+		cs.mu.Unlock()
+		return
+	}
+	cs.finished = true
+	cs.mu.Unlock()
+	// TODO(retry): commit current attempt if necessary.
+	cs.attempt.finish(err)
+	for _, o := range cs.opts {
+		o.after(cs.c)
+	}
+	cs.cancel()
+}
+
+func (a *csAttempt) context() context.Context {
+	return a.s.Context()
+}
+
+func (a *csAttempt) header() (metadata.MD, error) {
+	return a.s.Header()
+}
+
+func (a *csAttempt) trailer() metadata.MD {
+	return a.s.Trailer()
+}
+
+func (a *csAttempt) sendMsg(m interface{}) (err error) {
 	// TODO Investigate how to signal the stats handling party.
 	// generate error stats if err != nil && err != io.EOF?
+	cs := a.cs
 	defer func() {
 		// For non-client-streaming RPCs, we return nil instead of EOF on success
 		// because the generated code requires it.  finish is not called; RecvMsg()
@@ -381,14 +430,23 @@ func (cs *clientStream) SendMsg(m interface{}) (err error) {
 			err = nil
 		}
 		if err != nil && err != io.EOF {
-			// Call finish for errors generated by this SendMsg call.  (Transport
+			// Call finish on the client stream for errors generated by this SendMsg
+			// call, as these indicate problems created by this client.  (Transport
 			// errors are converted to an io.EOF error below; the real error will be
-			// returned from RecvMsg eventually in that case.)
+			// returned from RecvMsg eventually in that case, or be retried.)
 			cs.finish(err)
 		}
 	}()
+	// TODO: Check cs.sentLast and error if we already ended the stream.
+	if EnableTracing {
+		a.mu.Lock()
+		if a.trInfo.tr != nil {
+			a.trInfo.tr.LazyLog(&payload{sent: true, msg: m}, true)
+		}
+		a.mu.Unlock()
+	}
 	var outPayload *stats.OutPayload
-	if cs.statsHandler != nil {
+	if a.statsHandler != nil {
 		outPayload = &stats.OutPayload{
 			Client: true,
 		}
@@ -403,18 +461,19 @@ func (cs *clientStream) SendMsg(m interface{}) (err error) {
 	if !cs.desc.ClientStreams {
 		cs.sentLast = true
 	}
-	err = cs.t.Write(cs.s, hdr, data, &transport.Options{Last: !cs.desc.ClientStreams})
+	err = a.t.Write(a.s, hdr, data, &transport.Options{Last: !cs.desc.ClientStreams})
 	if err == nil {
 		if outPayload != nil {
 			outPayload.SentTime = time.Now()
-			cs.statsHandler.HandleRPC(cs.statsCtx, outPayload)
+			a.statsHandler.HandleRPC(a.ctx, outPayload)
 		}
 		return nil
 	}
 	return io.EOF
 }
 
-func (cs *clientStream) RecvMsg(m interface{}) (err error) {
+func (a *csAttempt) recvMsg(m interface{}) (err error) {
+	cs := a.cs
 	defer func() {
 		if err != nil || !cs.desc.ServerStreams {
 			// err != nil or non-server-streaming indicates end of stream.
@@ -422,46 +481,46 @@ func (cs *clientStream) RecvMsg(m interface{}) (err error) {
 		}
 	}()
 	var inPayload *stats.InPayload
-	if cs.statsHandler != nil {
+	if a.statsHandler != nil {
 		inPayload = &stats.InPayload{
 			Client: true,
 		}
 	}
-	if !cs.decompSet {
+	if !a.decompSet {
 		// Block until we receive headers containing received message encoding.
-		if ct := cs.s.RecvCompress(); ct != "" && ct != encoding.Identity {
+		if ct := a.s.RecvCompress(); ct != "" && ct != encoding.Identity {
-			if cs.dc == nil || cs.dc.Type() != ct {
+			if a.dc == nil || a.dc.Type() != ct {
 				// No configured decompressor, or it does not match the incoming
 				// message encoding; attempt to find a registered compressor that does.
-				cs.dc = nil
+				a.dc = nil
-				cs.decomp = encoding.GetCompressor(ct)
+				a.decomp = encoding.GetCompressor(ct)
 			}
 		} else {
 			// No compression is used; disable our decompressor.
-			cs.dc = nil
+			a.dc = nil
 		}
 		// Only initialize this state once per stream.
-		cs.decompSet = true
+		a.decompSet = true
 	}
-	err = recv(cs.p, cs.codec, cs.s, cs.dc, m, *cs.c.maxReceiveMessageSize, inPayload, cs.decomp)
+	err = recv(a.p, cs.codec, a.s, a.dc, m, *cs.c.maxReceiveMessageSize, inPayload, a.decomp)
 	if err != nil {
 		if err == io.EOF {
-			if statusErr := cs.s.Status().Err(); statusErr != nil {
+			if statusErr := a.s.Status().Err(); statusErr != nil {
 				return statusErr
 			}
 			return io.EOF // indicates successful end of stream.
 		}
 		return toRPCErr(err)
 	}
-	if cs.tracing {
+	if EnableTracing {
-		cs.mu.Lock()
+		a.mu.Lock()
-		if cs.trInfo.tr != nil {
+		if a.trInfo.tr != nil {
-			cs.trInfo.tr.LazyLog(&payload{sent: false, msg: m}, true)
+			a.trInfo.tr.LazyLog(&payload{sent: false, msg: m}, true)
 		}
-		cs.mu.Unlock()
+		a.mu.Unlock()
 	}
 	if inPayload != nil {
-		cs.statsHandler.HandleRPC(cs.statsCtx, inPayload)
+		a.statsHandler.HandleRPC(a.ctx, inPayload)
 	}
 	if cs.desc.ServerStreams {
 		// Subsequent messages should be received by subsequent RecvMsg calls.
@@ -470,75 +529,59 @@ func (cs *clientStream) RecvMsg(m interface{}) (err error) {
 
 	// Special handling for non-server-stream rpcs.
 	// This recv expects EOF or errors, so we don't collect inPayload.
-	err = recv(cs.p, cs.codec, cs.s, cs.dc, m, *cs.c.maxReceiveMessageSize, nil, cs.decomp)
+	err = recv(a.p, cs.codec, a.s, a.dc, m, *cs.c.maxReceiveMessageSize, nil, a.decomp)
 	if err == nil {
 		return toRPCErr(errors.New("grpc: client streaming protocol violation: get <nil>, want <EOF>"))
 	}
 	if err == io.EOF {
-		return cs.s.Status().Err() // non-server streaming Recv returns nil on success
+		return a.s.Status().Err() // non-server streaming Recv returns nil on success
 	}
 	return toRPCErr(err)
 }
 
-func (cs *clientStream) CloseSend() error {
+func (a *csAttempt) closeSend() {
+	cs := a.cs
 	if cs.sentLast {
-		return nil
-	}
-	cs.sentLast = true
-	cs.t.Write(cs.s, nil, nil, &transport.Options{Last: true})
-	// We ignore errors from Write and always return nil here.  Any error it
-	// would return would also be returned by a subsequent RecvMsg call, and the
-	// user is supposed to always finish the stream by calling RecvMsg until it
-	// returns err != nil.
-	return nil
-}
-
-func (cs *clientStream) finish(err error) {
-	if err == io.EOF {
-		// Ending a stream with EOF indicates a success.
-		err = nil
-	}
-	cs.mu.Lock()
-	defer cs.mu.Unlock()
-	if cs.finished {
 		return
 	}
-	cs.finished = true
+	cs.sentLast = true
-	cs.t.CloseStream(cs.s, err)
+	cs.attempt.t.Write(cs.attempt.s, nil, nil, &transport.Options{Last: true})
-	for _, o := range cs.opts {
+	// We ignore errors from Write.  Any error it would return would also be
-		o.after(cs.c)
+	// returned by a subsequent RecvMsg call, and the user is supposed to always
-	}
+	// finish the stream by calling RecvMsg until it returns err != nil.
-	if cs.done != nil {
+}
-		cs.done(balancer.DoneInfo{
+
+func (a *csAttempt) finish(err error) {
+	a.mu.Lock()
+	a.t.CloseStream(a.s, err)
+
+	if a.done != nil {
+		a.done(balancer.DoneInfo{
 			Err:           err,
 			BytesSent:     true,
-			BytesReceived: cs.s.BytesReceived(),
+			BytesReceived: a.s.BytesReceived(),
 		})
-		cs.done = nil
 	}
-	if cs.statsHandler != nil {
+	if a.statsHandler != nil {
 		end := &stats.End{
 			Client:    true,
-			BeginTime: cs.beginTime,
+			BeginTime: a.beginTime,
 			EndTime:   time.Now(),
 			Error:     err,
 		}
-		cs.statsHandler.HandleRPC(cs.statsCtx, end)
+		a.statsHandler.HandleRPC(a.ctx, end)
 	}
-	cs.cancel()
+	if a.trInfo.tr != nil {
-	if !cs.tracing {
-		return
-	}
-	if cs.trInfo.tr != nil {
 		if err == nil {
-			cs.trInfo.tr.LazyPrintf("RPC: [OK]")
+			a.trInfo.tr.LazyPrintf("RPC: [OK]")
 		} else {
-			cs.trInfo.tr.LazyPrintf("RPC: [%v]", err)
+			a.trInfo.tr.LazyPrintf("RPC: [%v]", err)
-			cs.trInfo.tr.SetError()
+			a.trInfo.tr.SetError()
 		}
-		cs.trInfo.tr.Finish()
+		a.trInfo.tr.Finish()
-		cs.trInfo.tr = nil
+		a.trInfo.tr = nil
 	}
+	a.mu.Unlock()
 }
 
 // ServerStream defines the interface a server stream has to satisfy.